Research On Electrical Submersible Pump System Operational Evaluation Based On Data Mining

According to the division standard of the development phases of oil fields, most onshore oil fields of China which have been put into production are getting in the middle-late development stage, with their declining production and high water cut. Electrical submersible pumps(ESP) are taking the lead, and playing a decisive and irreplaceable role in the current oil companies, because of its large displacement,high delivery lift and other unique advantages. However, in the operation and management process of the ESP unit, there exist some practical problems such as extensive management, block management, and lag of technology update, which directly affect the efficiency and operating cost of the ESP unit. The project of how to make submersible pump efficiently work, and reduce the production cost, is placed in front of the petroleum workers and engineers. In this paper, considering the inherent characteristics of ESP unit, and based on a comprehensive reference to the worldwide ESP technology, application and management experience, a systematic study on the theory, algorithm and experiment for the evaluation of ESP system has been done.First of all, the concept of comprehensive evaluation system for ESP system is proposed, and an evaluation model of ESP system based on Data Mining(DM) is established in order to improve the efficiency of the ESP and extend its service life.The comprehensive evaluation system for ESP system consists of three components,namely ESP economic benefit evaluation, ESP technical operation evaluation, and ESP reliability evaluation. Fuzzy evaluation method is used to establish the comprehensive evaluation model.Secondly, an automatic selection method is proposed for a DM model. Based on the characteristics of DM method and the actual demand of index screening for ESP system evaluation, clustering methodology is chosen among many DM algorithms,which can perform index screening for ESP system evaluation. Applying the variable selection R clustering analysis method, 15 economic evaluation indexes, 22 technical evaluation indexes and 19 reliability evaluation indexes are constructed. Based on the analytic hierarchy process(AHP), the reasonable weight value of evaluation indexes on all levels is determined.Thirdly, an comprehensive evaluation model of ESP system is established. An improvement method to the standard LM-BP neural network algorithm is put forward.Based on the improved LM-BP neural network algorithm, an economic evaluation model is established. Using principal component analysis and the classical regression analysis, technical evaluation model is established. Based on the AHP method and fuzzy evaluation method, considering the weight values of the reliability indexes of ESP system, a reliability evaluation model is established. According to the relationship matrix of the economic, technical, and reliability indexes and the weight vector, a comprehensive evaluation model of ESP system is established by applying the comprehensive evaluation method.Finally, practical analysis is performed for ESP evaluation system based on DM.The 150m3/d pumps are taken as the evaluation case, which are the most typical and widely used ESP units in oil fields. The 17220 records and 26 data fields of 10 ESP well B2-20-P41 in nearly four years are taken as the evaluation sample. Three specific evaluations, namely economic evaluation, technical evaluation and reliability evaluation are performed respectively. A comprehensive evaluation of ESP unit is given based on the specific evaluations. The result is in coincidence with the practical data of the well in economy, efficiency and reliability of the ESP unit.